DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators

Susmita Dey Manasi; Sachin S. Sapatnekar

doi:10.1145/3394885.3431539

DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators

Susmita Dey Manasi, Sachin S. Sapatnekar

Electrical and Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

9 Scopus citations

Abstract

Scheduling computations in each layer of a convolutional neural network on a deep learning (DL) accelerator involves a large number of choices, each of which involves a different set of memory reuse and memory access patterns. Since memory transactions are the primary bottleneck in DL acceleration, these choices can strongly impact the energy and throughput of the accelerator. This work proposes an optimization framework, DeepOpt, for general ASIC-based systolic hardware accelerators for layer-specific and hardware-specific scheduling strategy for each layer of a CNN to optimize energy and latency. Optimal hardware allocation significantly reduces execution cost as compared to generic static hardware resource allocation, e.g., improvements of up to 50 in the energy-delay product for VGG-16 and 41 for GoogleNet-v1.

Original language	English (US)
Title of host publication	Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	235-241
Number of pages	7
ISBN (Electronic)	9781450379991
DOIs	https://doi.org/10.1145/3394885.3431539
State	Published - Jan 18 2021
Event	26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021 - Virtual, Online, Japan Duration: Jan 18 2021 → Jan 21 2021

Publication series

Name	Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

Conference

Conference	26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021
Country/Territory	Japan
City	Virtual, Online
Period	1/18/21 → 1/21/21

Bibliographical note

Funding Information:
We thank Z. Wang and A. B. Kahng (UCSD) for helping in modeling SRAM area. This work is supported in part by NSF (CCF-1763761).

Publisher Copyright:
© 2021 Association for Computing Machinery.

Keywords

CNN
hardware accelerator
scheduling

Access

10.1145/3394885.3431539

OpenUrl availability

Full text

Cite this

Manasi, S. D., & Sapatnekar, S. S. (2021). DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators. In Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021 (pp. 235-241). Article 3431539 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/3394885.3431539

DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators. / Manasi, Susmita Dey; Sapatnekar, Sachin S.
Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021. Institute of Electrical and Electronics Engineers Inc., 2021. p. 235-241 3431539 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Manasi, SD & Sapatnekar, SS 2021, DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators. in Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021., 3431539, Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, Institute of Electrical and Electronics Engineers Inc., pp. 235-241, 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021, Virtual, Online, Japan, 1/18/21. https://doi.org/10.1145/3394885.3431539

Manasi SD, Sapatnekar SS. DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators. In Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021. Institute of Electrical and Electronics Engineers Inc. 2021. p. 235-241. 3431539. (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). doi: 10.1145/3394885.3431539

Manasi, Susmita Dey ; Sapatnekar, Sachin S. / DeepOpt : Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators. Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021. Institute of Electrical and Electronics Engineers Inc., 2021. pp. 235-241 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

@inproceedings{1916453f9c5545e2a742aaf981ff73cc,

title = "DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators",

abstract = "Scheduling computations in each layer of a convolutional neural network on a deep learning (DL) accelerator involves a large number of choices, each of which involves a different set of memory reuse and memory access patterns. Since memory transactions are the primary bottleneck in DL acceleration, these choices can strongly impact the energy and throughput of the accelerator. This work proposes an optimization framework, DeepOpt, for general ASIC-based systolic hardware accelerators for layer-specific and hardware-specific scheduling strategy for each layer of a CNN to optimize energy and latency. Optimal hardware allocation significantly reduces execution cost as compared to generic static hardware resource allocation, e.g., improvements of up to 50 in the energy-delay product for VGG-16 and 41 for GoogleNet-v1.",

keywords = "CNN, hardware accelerator, scheduling",

author = "Manasi, {Susmita Dey} and Sapatnekar, {Sachin S.}",

note = "Funding Information: We thank Z. Wang and A. B. Kahng (UCSD) for helping in modeling SRAM area. This work is supported in part by NSF (CCF-1763761). Publisher Copyright: {\textcopyright} 2021 Association for Computing Machinery.; 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021 ; Conference date: 18-01-2021 Through 21-01-2021",

year = "2021",

month = jan,

day = "18",

doi = "10.1145/3394885.3431539",

language = "English (US)",

series = "Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "235--241",

booktitle = "Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021",

}

TY - GEN

T1 - DeepOpt

T2 - 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021

AU - Manasi, Susmita Dey

AU - Sapatnekar, Sachin S.

N1 - Funding Information: We thank Z. Wang and A. B. Kahng (UCSD) for helping in modeling SRAM area. This work is supported in part by NSF (CCF-1763761). Publisher Copyright: © 2021 Association for Computing Machinery.

PY - 2021/1/18

Y1 - 2021/1/18

N2 - Scheduling computations in each layer of a convolutional neural network on a deep learning (DL) accelerator involves a large number of choices, each of which involves a different set of memory reuse and memory access patterns. Since memory transactions are the primary bottleneck in DL acceleration, these choices can strongly impact the energy and throughput of the accelerator. This work proposes an optimization framework, DeepOpt, for general ASIC-based systolic hardware accelerators for layer-specific and hardware-specific scheduling strategy for each layer of a CNN to optimize energy and latency. Optimal hardware allocation significantly reduces execution cost as compared to generic static hardware resource allocation, e.g., improvements of up to 50 in the energy-delay product for VGG-16 and 41 for GoogleNet-v1.

AB - Scheduling computations in each layer of a convolutional neural network on a deep learning (DL) accelerator involves a large number of choices, each of which involves a different set of memory reuse and memory access patterns. Since memory transactions are the primary bottleneck in DL acceleration, these choices can strongly impact the energy and throughput of the accelerator. This work proposes an optimization framework, DeepOpt, for general ASIC-based systolic hardware accelerators for layer-specific and hardware-specific scheduling strategy for each layer of a CNN to optimize energy and latency. Optimal hardware allocation significantly reduces execution cost as compared to generic static hardware resource allocation, e.g., improvements of up to 50 in the energy-delay product for VGG-16 and 41 for GoogleNet-v1.

KW - CNN

KW - hardware accelerator

KW - scheduling

UR - http://www.scopus.com/inward/record.url?scp=85100533732&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85100533732&partnerID=8YFLogxK

U2 - 10.1145/3394885.3431539

DO - 10.1145/3394885.3431539

M3 - Conference contribution

AN - SCOPUS:85100533732

T3 - Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

SP - 235

EP - 241

BT - Proceedings of the 26th Asia and South Pacific Design Automation Conference, ASP-DAC 2021

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 18 January 2021 through 21 January 2021

ER -

DeepOpt: Optimized Scheduling of CNNWorkloads for ASIC-based Systolic Deep Learning Accelerators

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this